Method and device for model training and information prediction

Abstract

The embodiment of the present application provides a model training and information prediction method and device, including: obtaining a sample image and reference feature points; using an information prediction model to obtain the predicted feature points and CONF in the sample image predict ; Calculate the LOSS of the predicted feature point relative to the reference feature point landmark ; use LOSS landmark Estimate CONF groudtruth ; calculate conf predict relative to CONF groudtruth LOSS confidence ;according to LOSS landmark and LOSS confidence , calculate LOSS total ;according to LOSS total Adjust the network parameters of the information prediction model to realize the training of the information prediction model. The model trained by applying the solution provided by the embodiment of the present application can predict the confidence of the feature points.

Description

technical field [0001] The present application relates to the technical field of information processing, in particular to a method and device for model training and information prediction. Background technique [0002] In recent years, artificial intelligence technology has been widely used in various fields. In the application process, feature point prediction is required to meet the needs of various fields. For example, in order to achieve face angle correction, face pose analysis, etc., it is necessary to predict the feature points of the face in the image, that is, the key points of the face. [0003] Taking the prediction of face key points in an image as an example, in the prior art, a face key point prediction model based on a convolutional neural network is usually used for face key point prediction. That is, the image that needs to predict the key points of the face is used as the input information of the above-mentioned model, and is input to the above-mentioned m...

AI Technical Summary

Problems solved by technology

[0004] However, in the prior art, the above-mentioned models often only output the predicted face key points, and do not output the confidence of the predicted face key points, making it difficult for users to understand the reliability of the predicted results.

Images